Avoid deleting the project environment directory if it is not a virtual environment

[zanieb]

Closes #7519

[zanieb]

We do something far more complex for uv venv which has a similar error case

uv/crates/uv-virtualenv/src/virtualenv.rs

Lines 88 to 122 in f895c40

	match location.metadata() {
	Ok(metadata) => {
	if metadata.is_file() {
	return Err(Error::Io(io::Error::new(
	io::ErrorKind::AlreadyExists,
	format!("File exists at `{}`", location.user_display()),
	)));
	} else if metadata.is_dir() {
	if allow_existing {
	debug!("Allowing existing directory");
	} else if location.join("pyvenv.cfg").is_file() {
	debug!("Removing existing directory");
	fs::remove_dir_all(location)?;
	fs::create_dir_all(location)?;
	} else if location
	.read_dir()
	.is_ok_and(|mut dir| dir.next().is_none())
	{
	debug!("Ignoring empty directory");
	} else {
	return Err(Error::Io(io::Error::new(
	io::ErrorKind::AlreadyExists,
	format!(
	"The directory `{}` exists, but it's not a virtualenv",
	location.user_display()
	),
	)));
	}
	}
	}
	Err(err) if err.kind() == io::ErrorKind::NotFound => {
	fs::create_dir_all(location)?;
	}
	Err(err) => return Err(Error::Io(err)),
	}

Should we split that out and replicate it? Is it necessary here?

[charliermarsh]

I do like the more explicit error handling like we have there.

[zanieb]

I think this also will be partially resolved by #7527, I ported at least some of these checks. I can revisit shared logic later too.

[zanieb]

Going to have to fight with Windows CI

[charliermarsh]

Should this part be catching a more targeted error?

[zanieb]

Naw we just care that it is not a virtual environment. I think this is addressed in #7527 anyway since I add a real error variant for invalid environments.